Next: Building Qualitative Event Models
Up: TrackingLow Level Vision,
Previous: 3D Modeling of Human
- Author(s):
- Abstract:
-
Existing object tracking algorithms generally use some form of local
optimisation, assuming that an object's position and shape change
smoothly over time. In some situations this assumption is not valid:
the trackable shape of an object may change discontinuously, for
example if it is the 2D silhouette of a 3D object.
In this paper we propose a novel method for modelling temporal shape
discontinuities explicitly. Allowable shapes are represented as a
union of (learned) bounded regions within a shape space. Discontinuous
shape changes are described in terms of transitions between these
regions. Transition probabilities are learned from training sequences
and stored in a Markov model. In this way we can create `wormholes' in
shape space. Tracking with such models is via an adaptation of the
CONDENSATION algorithm.
[Cover Page] [Table of Contents] [Author Index] [FAQ]
International Conference on Computer Vision '98
Bombay, India.
CDROM version produced at I I T, Bombay..